Log file Viewer

[07:35:28] <MrJones> I am having a bit of trouble getting CCACHE to work in a setting where I'm not directly calling pip/setuptools, due to pip's randomized --build directory

[07:35:33] <MrJones> can I somehow set an env var that will make pip always choose one fixed build directory when --build is not specified, but still clean it up afterwards?

[13:28:17] <MrJones> does that lack of response mean "no, there's no such option"? xD I suppose it is a pretty specific quirk of ccache that wouldn't really be relevant to anyone else

[14:00:41] <toad_polo> MrJones: I think lack of response means no one currently available in the room knows the answer. It's still early in the US, though.

[14:02:55] <toad_polo> Tenuously related issue: https://github.com/pypa/pip/issues/6276

[15:12:48] <ngoldbaum> MrJones: perhaps you want an editable install? or do you need to cache builds for CI or something like that?

[17:36:37] <MrJones> ngoldbaum: I have no control over the pip call

[17:37:04] <MrJones> I need to cache builds because it's a pretty big build... but yeah it's a CI-like environment

[17:37:27] <MrJones> this is for python-for-android, which builds like a really huge amount of packages to put together a self-contained python environment for an android app

[17:37:51] <MrJones> so I'm pretty much away from pip through layers of wrapper, but it would be extremely helpful to me if I could just make it stop randomizing the build path

[17:38:35] <MrJones> now the thing is I actually contribute to python-for-android, but this is a matter that is relatively minor in comparison to many others ("just" caching) and many of the pip calls also aren't directly in python-for-android but indirectly through external projects' setup scripts

[17:39:01] <MrJones> so even if I would try to fix that by addressing all pip calls and actually go in there and put in --build on each single one, that wouldn't really be a doable task

[17:39:15] <MrJones> so that's why I wondered if there is an environment variable that does it

[17:39:41] <MrJones> because this is making up huge chunks of the build (probably 60%ish) and takes up like 20 minutes of it, and if I could cache that I'd probably down to 3-4 minutes or something for that part

[17:41:58] <MrJones> I mean this might seem like something I should just care less about, but for debugging issues it just makes quite a productivy difference of having a build that has a turnaround time of 25minutes ish or let's say, maybe 10

[20:17:55] <ngoldbaum> odd

[20:18:12] <ngoldbaum> the y/n prompt for pip uninstall is not longer seeing when i hit "enter"

[20:21:16] <dstufft> MrJones: all pip command flags can be passed as environment variables fwiw

[20:21:48] <ngoldbaum> instead it writes ^M to the prompy

[20:21:53] <ngoldbaum> prompt

[20:22:23] <dstufft> ngoldbaum: I think I've seen this error on generic command line utilities

[20:22:39] <dstufft> IIRC I fix it by closing my terminal and reopening it

[20:22:52] <dstufft> I think it's your terminal getting into some wierd mode or state

[20:27:55] <ngoldbaum> fun!

[21:16:10] <ngoldbaum> dstufft: that fixed it thanks

[21:16:15] <ngoldbaum> terminals: who even knows

[21:43:42] <njs> ngoldbaum: see https://www.linusakesson.net/programming/tty/ , in particular the section "configuring the tty device"

[21:46:15] <njs> there's a whole set of translations from keyboard key -> magic that happen inside the kernel's tty layer, and they're all configurable in arcane ways... but in particular they control both things like "when do we actually pass data on to the application?" (do we wait for end-of-line?), and "which keys count as end-of-line?"